Authors: Siddhi Pramod Lande, Samruddhi Ravindra Alhat
Abstract: Recognizing emotions is very important for connecting human emotions with artificial intelligence. This study introduces Emotion Sense, a sophisticated real-time facial emotion recognition system utilizing deep learning and explainable AI (XAI). The suggested system uses a better MobileNetV3 architecture along with Coordinate Attention (CA) and Grad-CAM visualization to get high accuracy and make the results easy to understand. The model recognizes seven fundamental human emotions: happiness, sadness, anger, surprise, fear, disgust, and neutrality. The FER-2013 data set. Emotion Sense solves two big problems that traditional CNN-based models have by combining real-time performance with explainability. This makes it both accurate and clear. The experimental results show that it is 90.2% accurate and runs smoothly at 25 frames per second on CPU devices. This shows that it is useful for real-world applications like healthcare, education, and human-computer interaction. This research is unique because it uses a hybrid design that balances speed, accuracy, and interpretability while staying strong in different real-world situations
DOI: http://doi.org/10.5281/zenodo.17491066
